Why replacing your developers with AI is a business trap

Thinking about replacing some or all of your developers with AI to save on overhead? While AI-based, no code solutions might dazzle in demos, they rarely translate into real world ROI, costing your business money rather than saving it in the long run. 

While AI tools can perform adequately at a narrow slice of software development, your software engineers perform a wide range of mission-critical tasks, none of which can be replaced by these industry tools. And while vendors know this, it won’t be mentioned during their product pitches for one simple reason: survival.

AI vendors are desperate to achieve ROI

With all the AI hype right now, lots of investor money has been thrown at AI companies—and very soon, they’ve got to make that back, plus interest. 

Let’s pull back the wizard’s curtain and look at the math. Take OpenAI, for example. They’re the current market leader in Generative AI, and it seems like everyone in the world is using ChatGPT for everything from work emails to coding and cheating on homework. But how much are they actually making?

$13 billion in annual revenue, if you ask Sam Altman.

That’s not so bad for a tech start-up, right? But how much are they earning once you take into account capital expenditures?

Negative $145 billion. And they’ve committed to spending $1 trillion on infrastructure over the next decade.

To quote Sesame Street, one of these things doesn’t belong. OpenAI has no plans to be profitable before 2030, by which time it has told investors the company will be making $200 billion a year—a whopping x15 what they’re earning now. Meanwhile, ChatGPT subscriber growth has stalled according to multiple industry reports.

Figures like this hang like a Sword of Damocles over sales and marketing teams, who have to fix this shortfall. Customers need to buy, buy, buy. And that’s where your company comes in.

The million-dollar developer with an AI target on their back

This morning, I went out for coffee with a friend. As the conversation swayed towards developer salaries, they told me there had been a cybersecurity incident at their company, and they had to mobilize a highly paid specialist to get involved. 

“One of these developers gets paid a million dollars!” they breathlessly exclaimed, with a tone that they clearly thought this was too much. As someone in leadership, they only made a fraction of that.

It’s not the first time I’ve heard this sentiment: that tech professionals—normally software engineers—are entitled, overpaid, and difficult to work with, and have no right being paid so much as an individual contributor. As Corey Doctorow put it, developers are the “Princes of Labor,” possessing enviable skills that allow them to be courted in a way unlike other career professionals.

These sizable wages are also big, annoying dents in a company’s financial metrics. AI companies know this, and that’s why their current pitch—both to investors and potential clients—is all about AI job replacement. The math is very simple: fire that highly paid developer, save yourself most of that money, and pay us a small fraction of that cost. 

Of course, I’ll point out that only a small number of devs actually get paid a million dollars. The real average is around the $200k+ mark, but it’s still nothing to be sneezed at, and for a development team, this cost adds up. To any company with a development team, cutting these costs is an attractive proposition. 

It’s also a supremely bad move. 

AI can’t do the job of a skilled developer; not even close

…Which is not news to developers, but often is to everyone else.

Developers are often well-paid because what they do is mysterious and difficult to understand to non-devs. Unfortunately, this is a double-edged sword: because it’s hard to understand what devs do, AI salespeople can convince business leaders that their AI tool can do the same thing, especially when they present some pretty code and a vibe-coded demo product.

The truth is that AI can’t do a developer’s job. But the truth isn’t important. Why? The decision to replace a human with an AI isn’t about whether or not the AI is up to the task. It’s about whether or not the employer believes the AI can do that job.

AI vendors have a vested interest in selling you this idea. They’re under pressure to pay off all those dizzyingly large investment figures. Once you’ve fired 90% of your developers and replaced them with AI tools, they’ve won. Your company is in the vendor lock-in zone. It would be nice if you didn’t crash and burn (for case studies) but it keeps the Sword of Damocles at bay a little longer. 

Unfortunately, the only one who can battle this information is the developer, who’s also the most at threat. To leaders, their warnings sound like a chicken warning a fox that they taste bad. 

“Of course they’d say AI can’t do their job!” they say. “They’re going to lose their cushy, overpaid position.”

Why AI can’t do a skilled developer’s job

What AI vendors tend to pitch as miracle developer replacements are typically coding solutions, but coding is only a fraction of most developer’s jobs, somewhere in the range of 20%. 

The vast majority of their work is actually solving problems. When someone says “Make a thing,” it’s the engineer who asks questions like “How do we make this thing not conflict with the last thing?” “How will users interact with this thing?” “Does this thing actually do something we want?” And so on. In short, requirements gathering and smart architecture. They typically spend longer trying to gather information on what to develop to meet business needs than it takes to actually build the solution. Measure twice, cut once.

AI can’t answer these questions. It can build you a solution fast, but it won’t be the solution. 

That takes practitioner knowledge, logic, and soft skills. Giving the AI tool access to your existing codebase or training your non-technical staff in better prompting doesn’t provide that. 

Without an expert guiding it, the solutions that an AI produces tend to create increasing technical debt. Let’s say you use it to release an app to sell. The initial release might be fine, but have some bugs, so you ask AI to fix the bugs. But in the process, it rewrites key chunks of the whole program, breaking issues for your users. Same thing for when you try to expand the product with new features. 

Why? It has no ability to “think” in an architectural way. When you ask it to fix a wheel, it will reinvent it, giving you square versions when the rest are round, or a smaller one than the rest, or one with not enough spokes to hold the weight of the vehicle. 

Generative AI is really designed for repetitive, low-risk tasks where accuracy and quality are not big factors. For example, summarizing a Zoom meeting is pretty low risk, and if there’s a few mistakes, it’s not a big deal. Meanwhile, booking surgery in a calendar might be repetitive, but accuracy is of extreme importance. 

Your software is a place where accuracy and quality are both very important, and a quality dip can have damaging effects on your business, from slowing delivery to introducing damaging compliance and security issues.

Replacing junior devs has short term gains, but long term pains

According to Pluralsight’s latest Tech Forecast, industry experts are seeing quite a few businesses replace junior developers with AI tools. While AI tools are cheaper than another salary, even at the entry level, this is another business trap. This one is slower to snap, but obvious when you think about it.

Because they’re not immortal, eventually your experienced developers move on or retire. Without junior developers to become mid-level and then senior developers, you’ve got to hire externally to fill these gaps. 

Hiring for senior dev positions is already hard—that’s why they have so much bargaining power—and it’s only getting worse. Nearly nine in ten organizations (89%) now say hiring is more expensive than upskilling for IT roles, according to Pluralsight’s latest Tech Skills Report. That’s up from nearly half (49%) the year before, a sizable jump. 

That’s why you’re often better off upskilling rather than hiring. But even then, you need those junior positions for people to move into—you can’t give someone access to an AI tool and expert them to go straight into mid or senior-level development. 

Conclusion: Your need for skilled developers isn’t going anywhere

Beware the hype of anyone telling you otherwise. Your tech stack is typically either complex and requires knowledgeable developers to make things happen, or it’s blissfully simple, and requires those same professionals to stay that way. Meanwhile, you need to keep investing in new talent to replace that critical talent when it inevitably leaves.

Yes, it would be nice to have millions of dollars extra on the spreadsheet. But you’re paying those developers to do an irreplaceable job—one where AI is simply not up to the task.